Joint source-decoding in large scale sensor networks using Markov random field models

An approach to scalable joint source decoding in large-scale sensor networks, based on Markov-random filed (MRF) modeling of the spatio-temporal correlation in the observations is presented. This approach exploits the correlation among a multitude of sensors for joint decoding at a central decoder, while using simple distributed quantizers in individual sensors. The decoder derivations are provided for Slepian–Wolf coded quantization based on both sample-by-sample (scalar) binning and vector binning schemes constructed via channel code partitioning. Simulation results are presented to demonstrate the performance achievable with the proposed decoding approach.